CERBERUS/LibreChat
1
0
Fork 0
mirror of https://github.com/danny-avila/LibreChat.git synced 2026-06-09 17:31:19 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 6

🧮 chore: Update Gemma Context Token Defaults (#13410)

Browse source
This commit is contained in:
Danny Avila 2026-05-30 00:29:19 -04:00 • committed by GitHub
parent a16f08a425
commit cb6bd71ab9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 41 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

31
api/utils/tokens.spec.js
View file

@ -356,6 +356,37 @@ describe('getModelMaxTokens', () => {
);
});
test('should return correct context tokens for Gemma models', () => {
expect(maxTokensMap[EModelEndpoint.google].gemma).toBe(32768);
expect(getModelMaxTokens('gemma', EModelEndpoint.google)).toBe(
maxTokensMap[EModelEndpoint.google].gemma,
);
expect(getModelMaxTokens('gemma-2-9b-it', EModelEndpoint.google)).toBe(
maxTokensMap[EModelEndpoint.google]['gemma-2'],
);
expect(getModelMaxTokens('gemma-3-27b-it', EModelEndpoint.google)).toBe(
maxTokensMap[EModelEndpoint.google]['gemma-3-27b'],
);
expect(getModelMaxTokens('gemma4:latest', EModelEndpoint.google)).toBe(
maxTokensMap[EModelEndpoint.google].gemma4,
);
expect(getModelMaxTokens('gemma4:e4b', EModelEndpoint.google)).toBe(
maxTokensMap[EModelEndpoint.google].gemma4,
);
expect(getModelMaxTokens('Gemma4:31B', EModelEndpoint.custom)).toBe(
maxTokensMap[EModelEndpoint.custom]['gemma4:31b'],
);
expect(getModelMaxTokens('ollama/gemma4:31b', EModelEndpoint.custom)).toBe(
maxTokensMap[EModelEndpoint.custom]['gemma4:31b'],
);
expect(getModelMaxTokens('google/gemma-4-31B-it', EModelEndpoint.google)).toBe(
maxTokensMap[EModelEndpoint.google]['gemma-4-31b'],
);
expect(getModelMaxTokens('google/gemma-4-26B-A4B-it', EModelEndpoint.google)).toBe(
maxTokensMap[EModelEndpoint.google]['gemma-4-26b-a4b'],
);
});
test('should return correct tokens for partial match - Cohere models', () => {
expect(getModelMaxTokens('command', EModelEndpoint.custom)).toBe(
maxTokensMap[EModelEndpoint.custom]['command'],

11
packages/api/src/utils/tokens.ts
View file

@ -95,10 +95,19 @@ const cohereModels = {
const googleModels = {
/* Max I/O is combined so we subtract the amount from max response tokens for actual total */
gemma: 8196,
gemma: 32768,
'gemma-2': 32768,
'gemma-3': 32768,
'gemma-3-27b': 131072,
'gemma4:31b': 256000,
'gemma4-31b': 256000,
'gemma-4-31b': 256000,
'gemma4:26b': 256000,
'gemma4-26b': 256000,
'gemma-4-26b-a4b': 256000,
'gemma-4-26b': 256000,
gemma4: 128000,
'gemma-4': 128000,
gemini: 30720, // -2048 from max
'gemini-pro-vision': 12288,
'gemini-1.5': 1000000,